• 0 Posts
  • 49 Comments
Joined 1 year ago
cake
Cake day: October 26th, 2023

help-circle
  • This why you check your backups periodically and replace the bad ones with good copies. If you’re asking how you know what’s good and bad - traditionally and fundamentally, even if many people here dismiss it, the storage already has checksums, that sneaky bitrot when the storage will give you slightly altered data (instead of saying “Error”) are so small that most people would never encounter this. Now of course serious data hoarders would use checksumming file systems, will do extra checksums for any archived data, also all archiving formats or backup formats have their own checksums too, if one would use that instead of dropping the files in the regular file system.




  • Nextcloud server has 907 contributors, and 2672 Open issues and 37970 Closed. That is without the desktop client, the mobile apps (multiple), and everything around (viewers, editors, voice/video/IM app, and so on. It’s packaged in 101 different ways, enough of them being all kinds of one command/click/etc. install, and you can even buy it directly managed and hosted and everything from multiple providers, including (but not limited) from NextCloud themselves. I think it’s extremely hard to compete in any way, even if you have something much simpler.



  • I’d do directly Wireguard (this is what Tailscale uses but it’s clear and controllable instead of more automagical). Openvpn or even directly openssh (of course configured with pubkeys) would be similar (run everything on non-standard ports to keep things quieter, and a non-standard user if applicable).

    You can do local encryption there too, with LUKS, zfs, ecryptfs or even rclone (actually you can do it locally with rclone so the remote never sees cleartext).











  • I’ve read that there is 1 (or 2) parity drives, but does this mean that these drives always spin up with everything you do? Wouldn’t that shorten its lifespan significantly compared to a SHR configuration?

    SHR is mdadm RAID. It means:

    • you can lose more data than the drives you’ve lost
    • you MUST to spin up ALL drives for mostly everything
    • you need all the disks as well in some kind of service recovery situation (like for example you still have all the disks fine but your box died, you need all 8 or whatever number of disks to be connected to do anything, can’t just take one disk and use the data from there)

    If you WANT to spin all drives all the time with unraid of course you CAN do that, but otherwise you CAN spin up only the one from where you read (or the one you write to + the parity). That is of course an advantage, you can do as you wish.


  • What isn’t shown in the chart is the price for larger drives. We’ve had prices in the thousands (discussing dollars), then in the hundreds, then the HUGE difference was made from the crisis started at the end of 2011, when we started with 2TBs going towards 25/TB (ok, usually 30+ but in any case way under 100 for a 2TB drive). The prices went up and then barely recovered. Not only the maximum capacity didn’t increase too fast but also the price per TB didn’t decrease much. We aren’t nitpicking here that it’s between 25/TB and 15/TB or 35 and 12.5, we’re talking orders of magnitude like it was the case for these intervals before. There isn’t anything worth mentioning into double digits anymore (spinning drives under $100). And it isn’t that there wasn’t inflation in the 80s, 90s and 2000s, this thing with nearly zero inflation was just some part of 2010s.

    There are people even now (well, as of yesterday) in this sub that would recommend 4TB spinning drives for 135 Euros. That is when you can have SSDs at around 150 with the right sale (a little under even if accept Samsung’s QLC, which is fine).